Jakub – Golang, Kubernetes, Temporal

A TMF-compliant platform for telecommunication system management. Complex business processes are modelled as long-running workflows that orchestrate work across many systems, including legacy services with little or no documentation. Built in Go on Temporal and Kafka and deployed on Kubernetes, with cluster operations handled by a separate platform team.

• Created a DSL that lets other users define business-process workflows, which the platform then runs on Temporal;
• Designed workflows to be resilient: when a step fails, the workflow pauses, waits for an automated or manual fix, validates it and continues, so the system can self-heal;
• Introduced the actor model to workflow design (one workflow per resource, with commands arriving as a message stream), which removed the race conditions seen with standard workflows;
• Built integrations with multiple legacy systems with very little information from other parties, using OpenAPI schemas and generated models that are refined as the real behavior of each service becomes clear;
• Set up observability for these integrations with logging HTTP clients and alerts, and handled authentication through OAuth behind an API gateway, with credentials kept as Kubernetes secrets and configuration in ConfigMaps;
• Uses AI coding assistants (Claude Code) daily while reviewing the generated code to keep a solid mental model of the system.

Conversational voicebot platform. Built Rust services that combined synchronous APIs with asynchronous messaging (Kafka, Pub/Sub) on Google Cloud.

• Built the real-time pipeline for voicebot calls: a live audio stream coming from Twilio was passed to a transcription service, then to an LLM and back, while the same stream was also stored in Google Cloud Storage and transformed for later analytics;
• Owned a couple of services end to end, designing new ones or extending existing ones in team design sessions;
• Modelled domains in Rust so that invalid states cannot be represented;
• Learned Rust and became productive in it within a month of moving to the project;
• Wrote Python base code as a template for teammates when the company decided to switch from Rust to Python.

Cloud backup service built in Go. The backup program needed nodes with enough capacity to run, so the service provisioned them on demand in the cloud and shut them down as soon as each backup completed, which kept infrastructure costs down.

• Managed the service that provisioned cloud nodes with the capacity required to run backup jobs and decommissioned them after the backup completed, to cut infrastructure costs;
• Worked on a small set of services, taking part in design sessions to decide whether to create a new service or extend an existing one;
• Combined synchronous APIs with asynchronous messaging (Kafka, Pub/Sub) between services.

A conversational intelligence platform that analyzes call center conversations, handling a high volume of recorded and live data.

• Built the storage and search layer for conversation analysis on PostgreSQL and Apache Solr, indexing a large volume of conversations;
• Integrated a biometric AI service into the company's main product;
• Extended the product with voice footprint comparison;
• Managed live data gathering from TV and radio;
• Integrated multiple services into the main product.